dynamicfusion
dynamicfusion copied to clipboard
mihaibujanca
install with cmake, windows
Got the error message as following. Do you have any guide for building with cmake+VS2015. Thanks.
CMake Error at C:/Program Files/CMake/share/cmake-3.10/Modules/FindPackageHandleStandardArgs.cmake:137 (message): Could NOT find OpenCV (missing: viz) (found version "3.3.0") Call Stack (most recent call first): C:/Program Files/CMake/share/cmake-3.10/Modules/FindPackageHandleStandardArgs.cmake:378 (_FPHSA_FAILURE_MESSAGE) C:/Users/dengya/Documents/opencv_files/opencv/build/x64/vc14/lib/OpenCVConfig.cmake:261 (find_package_handle_standard_args) C:/Users/dengya/Documents/opencv_files/opencv/build/OpenCVConfig.cmake:122 (include) CMakeLists.txt:47 (find_package)
Need to recmake and rebuild your opencv project manually. Your opencv lack the viz module.
Thank you. Have you installed in windows using cmake and vs? I have issue to build google-glog and suitesparse with cmake. Have you successfully build that? Ceres is not listed in the windows version's dependencies. However, when I tried to build the dynamic fusion, it said ceres is not found. Therefore, I think it does need cere pre-installed.
On Mon, Nov 20, 2017 at 11:08 PM, HaoguangHuang [email protected] wrote:
Need to recmake and rebuild your opencv project manually. Your opencv lack the viz module.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/mihaibujanca/dynamicfusion/issues/33#issuecomment-345936423, or mute the thread https://github.com/notifications/unsubscribe-auth/AH0KC0spQTViKS8yjAu2ezjwyvpaXHOjks5s4nbjgaJpZM4QlCbp .
I have just built the project dynamicfusion successfully on my ubuntu 16.04, but haven't tried on windows.
The latest version of this project seems to use Opt instead of Ceres, to process optimization of warp field.
I will need to update the documentation and building process. Probably going to find some time for it this weekend.
Indeed the current version is using Opt rather than Ceres, so you should be able to work around installing Ceres, but I don't think that the code is modular enough at the moment to let you just comment out the dependency in the CMakeLists
@HaoguangHuang Any chance you could write what was needed other than the current instructions to get it working on Windows?
I'll need to write this up for other users, would appreciate any suggestions on that
Hi,
Built the ceres and another error occurred when I tried to build dynamic fusion. Can I disable the use of OPENNI? I thought that's optional. Thank you.
CMake Error: The following variables are used in this project, but they are set to NOTFOUND. Please set them or make sure they are set and tested correctly in the CMake files: OPENNI_INCLUDE_DIR (ADVANCED) used as include directory in directory C:/Users/dengya/Documents/dynamicfusion used as include directory in directory C:/Users/dengya/Documents/dynamicfusion/kfusion used as include directory in directory C:/Users/dengya/Documents/dynamicfusion/kfusion used as include directory in directory C:/Users/dengya/Documents/dynamicfusion/apps used as include directory in directory C:/Users/dengya/Documents/dynamicfusion/apps OPENNI_LIBRARY (ADVANCED) linked by target "kfusion" in directory C:/Users/dengya/Documents/dynamicfusion/kfusion
On Wed, Nov 22, 2017 at 8:38 AM, Mihai [email protected] wrote:
@HaoguangHuang https://github.com/haoguanghuang Any chance you could write what was needed other than the current instructions to get it working on Windows? I'll need to write this up for other users, would appreciate any suggestions on that
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/mihaibujanca/dynamicfusion/issues/33#issuecomment-346405663, or mute the thread https://github.com/notifications/unsubscribe-auth/AH0KCz_UeKWFeC43hV4VzUh7ZD7xRZ1Tks5s5E3wgaJpZM4QlCbp .
@mihaibujanca If you need, I am pleased to supple some simple instructions here about how to build this project on ubuntu 16.04 soon.
@mxgbs The solution contains two projects that can produce executable file, one is dynamicfusion, the other is dynamicfusion_kinect. And they correspond to demo.cpp and dynamicfusion_kinect.cpp respectively. If you have no OPENNI installed, don't choose ALL_BUILD when build the project. Just building 'dynamicfusion' is enough
Hi Haoguang,
Thanks for your kindly reply. Do you have any idea how can I disable the dynamicfusion_kinect part and openNI dependency in CMakeLists.txt? Thank you.
Sincerely,
Yan
On Wed, Nov 22, 2017 at 8:12 PM, HaoguangHuang [email protected] wrote:
@mxgbs https://github.com/mxgbs The solution contains two projects that can produce executable file, one is dynamicfusion, the other is dynamicfusion_kinect. And they correspond to demo.cpp and dynamicfusion_kinect.cpp respectively. If you have no OPENNI installed, don't choose ALL_BUILD when build the project. Just building 'dynamicfusion' is enough
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/mihaibujanca/dynamicfusion/issues/33#issuecomment-346528747, or mute the thread https://github.com/notifications/unsubscribe-auth/AH0KCxlkUiL3tUzq07oc8PYOZgEnU90Uks5s5PDHgaJpZM4QlCbp .
One way is to just remove lines 18-27 from apps/CMakeLists.txt. If OpenNI is not installed, that shouldn't be built anyway - not sure why it appears to be a problem.
Alternatively just explicitly use
set(OPENNI_FOUND FALSE) somewhere before add_subdirectory(apps) in the main CMakeLists.txt
Hi Mihai,
Thanks for your quick response. I tried, and use cmake, the same error still occurs:. I think the reason may be because the following functions using some OPENNI functions? CMake Error: The following variables are used in this project, but they are set to NOTFOUND. Please set them or make sure they are set and tested correctly in the CMake files: OPENNI_INCLUDE_DIR (ADVANCED) used as include directory in directory C:/Users/dengya/Documents/dynamicfusion used as include directory in directory C:/Users/dengya/Documents/dynamicfusion/kfusion used as include directory in directory C:/Users/dengya/Documents/dynamicfusion/kfusion used as include directory in directory C:/Users/dengya/Documents/dynamicfusion/apps used as include directory in directory C:/Users/dengya/Documents/dynamicfusion/apps OPENNI_LIBRARY (ADVANCED) linked by target "kfusion" in directory C:/Users/dengya/Documents/dynamicfusion/kfusion
Sincerely,
Yan
On Mon, Nov 27, 2017 at 1:05 PM, Mihai [email protected] wrote:
One way is to just remove lines 18-27 from apps/CMakeLists.txt. If OpenNI is not installed, that shouldn't be built anyway - not sure why it appears to be a problem.
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/mihaibujanca/dynamicfusion/issues/33#issuecomment-347327663, or mute the thread https://github.com/notifications/unsubscribe-auth/AH0KC3y53O1y7XiY6vMm-NGAJRhklfgXks5s6yQUgaJpZM4QlCbp .
I removed dynamicfusion_kinect from apps. The .sln is successfully generated via cmake. However, new error as following occurs when building with VS2015. Trying to solve that now.
1>------ Build started: Project: ZERO_CHECK, Configuration: Release x64
1> Checking Build System 1> CMake does not need to re-run because C:/Users/dengya/Documents/dynamicfusion/build_new/CMakeFiles/generate.stamp is up-to-date. 1> CMake does not need to re-run because C:/Users/dengya/Documents/dynamicfusion/build_new/kfusion/CMakeFiles/generate.stamp is up-to-date. 1> CMake does not need to re-run because C:/Users/dengya/Documents/dynamicfusion/build_new/apps/CMakeFiles/generate.stamp is up-to-date. 2>------ Build started: Project: kfusion, Configuration: Release x64 ------ 2> Building NVCC (Device) object kfusion/CMakeFiles/kfusion.dir/src/cuda/Release/kfusion_generated_imgproc.cu.obj 2> imgproc.cu 2>cl : Command line error D8021: invalid numeric argument '/Wno-write-strings'
On Mon, Nov 27, 2017 at 1:17 PM, yan deng [email protected] wrote:
Hi Mihai,
Thanks for your quick response. I tried, and use cmake, the same error still occurs:. I think the reason may be because the following functions using some OPENNI functions? CMake Error: The following variables are used in this project, but they are set to NOTFOUND. Please set them or make sure they are set and tested correctly in the CMake files: OPENNI_INCLUDE_DIR (ADVANCED) used as include directory in directory C:/Users/dengya/Documents/ dynamicfusion used as include directory in directory C:/Users/dengya/Documents/ dynamicfusion/kfusion used as include directory in directory C:/Users/dengya/Documents/ dynamicfusion/kfusion used as include directory in directory C:/Users/dengya/Documents/ dynamicfusion/apps used as include directory in directory C:/Users/dengya/Documents/ dynamicfusion/apps OPENNI_LIBRARY (ADVANCED) linked by target "kfusion" in directory C:/Users/dengya/Documents/ dynamicfusion/kfusion
Sincerely,
Yan
On Mon, Nov 27, 2017 at 1:05 PM, Mihai [email protected] wrote:
One way is to just remove lines 18-27 from apps/CMakeLists.txt. If OpenNI is not installed, that shouldn't be built anyway - not sure why it appears to be a problem.
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/mihaibujanca/dynamicfusion/issues/33#issuecomment-347327663, or mute the thread https://github.com/notifications/unsubscribe-auth/AH0KC3y53O1y7XiY6vMm-NGAJRhklfgXks5s6yQUgaJpZM4QlCbp .
More errors come. For the mlibCore.h, should we install opt first and put that to the cmakelist? seems all the following libraries are missing. #include <cudaUtil.h> #include <SolverIteration.h> #include <CombinedSolverParameters.h> #include <CombinedSolverBase.h> #include <OptGraph.h>
1>------ Build started: Project: ZERO_CHECK, Configuration: Release x64
2>------ Build started: Project: kfusion, Configuration: Release x64 ------ 2> Building NVCC (Device) object kfusion/CMakeFiles/kfusion.dir/src/cuda/Release/kfusion_generated_imgproc.cu.obj 2> imgproc.cu 2>CUSTOMBUILD : nvcc warning : The -std=c++11 flag is not supported with the configured host compiler. Flag will be ignored. 2>c:\users\dengya\documents\dynamicfusion\kfusion\src\cuda\temp_utils.hpp(585): warning : function "__ballot" 2> c:\program files\nvidia gpu computing toolkit\cuda\v9.0\include\sm_20_intrinsics.h(407): here was declared deprecated ("__ballot() is deprecated in favor of __ballot_sync() and may be removed in a future release (Use -Wno-deprecated-declarations to suppress this warning).") 2> 2>c:\users\dengya\documents\dynamicfusion\kfusion\src\cuda\temp_utils.hpp(597): warning : function "__all" 2> c:\program files\nvidia gpu computing toolkit\cuda\v9.0\include\device_atomic_functions.h(181): here was declared deprecated ("__all() is deprecated in favor of __all_sync() and may be removed in a future release (Use -Wno-deprecated-declarations to suppress this warning).") 2> 2>c:\users\dengya\documents\dynamicfusion\kfusion\src\cuda\device.hpp(54): error : identifier "__float2half_rn" is undefined 2> 2>c:\users\dengya\documents\dynamicfusion\kfusion\src\cuda\device.hpp(59): error : identifier "__half2float" is undefined 2> 2>c:\users\dengya\documents\dynamicfusion\kfusion\src\cuda\device.hpp(61): error : identifier "__half2float" is undefined 2> 2>C:/Users/dengya/Documents/dynamicfusion/kfusion/src/cuda/imgproc.cu(270): error : identifier "__float2half_rn" is undefined 2> 2> 4 errors detected in the compilation of "C:/Users/dengya/AppData/Local/Temp/tmpxft_000061a0_00000000-13_imgproc.compute_61.cpp1.ii". 2>CUSTOMBUILD : nvcc warning : The -std=c++11 flag is not supported with the configured host compiler. Flag will be ignored. 2> imgproc.cu 2> CMake Error at kfusion_generated_imgproc.cu.obj.Release.cmake:279 (message): 2> Error generating file 2> C:/Users/dengya/Documents/dynamicfusion/build_new/kfusion/CMakeFiles/kfusion.dir/src/cuda/Release/kfusion_generated_imgproc.cu.obj 2> 2> 3>------ Build started: Project: dynamicfusion, Configuration: Release x64
3> demo.cpp 3>C:\Users\dengya\Documents\dynamicfusion\kfusion\src\utils\dual_quaternion.hpp(17): warning C4305: 'return': truncation from 'double' to 'float' 3>c:\users\dengya\documents\dynamicfusion\kfusion\include\opt\mLibInclude.h(15): fatal error C1083: Cannot open include file: 'mLibCore.h': No such file or directory 4>------ Build started: Project: ALL_BUILD, Configuration: Release x64
5>------ Skipped Build: Project: INSTALL, Configuration: Release x64 ------ 5>Project not selected to build for this solution configuration ========== Build: 2 succeeded, 2 failed, 0 up-to-date, 1 skipped ==========
On Mon, Nov 27, 2017 at 1:32 PM, Mihai [email protected] wrote:
Try removing -Wno-write-strings on line 5 of CMakeLists.txt
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/mihaibujanca/dynamicfusion/issues/33#issuecomment-347335141, or mute the thread https://github.com/notifications/unsubscribe-auth/AH0KC6hGUP7vWsfM1hvKdCO_HkfkiLGvks5s6yp2gaJpZM4QlCbp .
yeah you have to install Opt (and its dependencies) and modify OPT_INCLUDE_DIRS and OPT_LIBRARY in CMakeLists
Do you have any clues how to fix following errors?
dynamicfusion\kfusion\src\cuda\device.hpp(54): error : identifier "__float2half_rn" is undefined dynamicfusion\kfusion\src\cuda\device.hpp(59): error : identifier "__half2float" is undefined dynamicfusion\kfusion\src\cuda\device.hpp(61): error : identifier "__half2float" is undefined dynamicfusion/kfusion/src/cuda/imgproc.cu(270): error : identifier "__float2half_rn" is undefined
Sincerely,
Yan
On Mon, Nov 27, 2017 at 1:48 PM, Mihai [email protected] wrote:
yeah you have to install Opt (and its dependencies) and modify OPT_INCLUDE_DIRS and OPT_LIBRARY in CMakeLists
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/mihaibujanca/dynamicfusion/issues/33#issuecomment-347339364, or mute the thread https://github.com/notifications/unsubscribe-auth/AH0KCwGjNaYlpbX4hA0K116cJ9_hubwIks5s6y4fgaJpZM4QlCbp .
Besides, is this the opt you used https://software.sandia.gov/opt++/ ?
Sincerely,
Yan
On Mon, Nov 27, 2017 at 1:50 PM, yan deng [email protected] wrote:
Do you have any clues how to fix following errors?
dynamicfusion\kfusion\src\cuda\device.hpp(54): error : identifier "__float2half_rn" is undefined dynamicfusion\kfusion\src\cuda\device.hpp(59): error : identifier "__half2float" is undefined dynamicfusion\kfusion\src\cuda\device.hpp(61): error : identifier "__half2float" is undefined dynamicfusion/kfusion/src/cuda/imgproc.cu(270): error : identifier "__float2half_rn" is undefined
Sincerely,
Yan
On Mon, Nov 27, 2017 at 1:48 PM, Mihai [email protected] wrote:
yeah you have to install Opt (and its dependencies) and modify OPT_INCLUDE_DIRS and OPT_LIBRARY in CMakeLists
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/mihaibujanca/dynamicfusion/issues/33#issuecomment-347339364, or mute the thread https://github.com/notifications/unsubscribe-auth/AH0KCwGjNaYlpbX4hA0K116cJ9_hubwIks5s6y4fgaJpZM4QlCbp .
The errors on undefined identifiers would come either from missing CUDA dependencies or incompatible CUDA version: http://docs.nvidia.com/cuda/cuda-math-api/group__CUDA__MATH__INTRINSIC__CAST.html
I use the latest CUDA version 9.0. Should I downgrade? First time using CUDA, lack of experience. Have you build this project on Windows successfully? If no one did, should I just give up and try other platform? I recalled you mentioned that you're preparing another tutorial for windows install, how is going? Thanks.
Sincerely,
Yan
On Mon, Nov 27, 2017 at 2:14 PM, Mihai [email protected] wrote:
The errors on undefined identifiers would come either from missing CUDA dependencies or incompatible CUDA version
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/mihaibujanca/dynamicfusion/issues/33#issuecomment-347346589, or mute the thread https://github.com/notifications/unsubscribe-auth/AH0KC9qiGNJAC37CjM0dZUmX2Ozc1S4qks5s6zREgaJpZM4QlCbp .
Hi Mihai,
I think the problem is here" Unsupported Features General CUDA ‣ CUDA library. The built-in functions __float2half_rn() and __half2float() have been removed. Use equivalent functionality in the updated fp16 header file from the CUDA toolkit." in https://developer.download.nvidia.com/compute/cuda/9.0/Prod/docs/sidebar/CUDA_Toolkit_Release_Notes.pdf
Sincerely,
Yan
On Tue, Nov 28, 2017 at 2:22 PM, yan deng [email protected] wrote:
I use the latest CUDA version 9.0. Should I downgrade? First time using CUDA, lack of experience. Have you build this project on Windows successfully? If no one did, should I just give up and try other platform? I recalled you mentioned that you're preparing another tutorial for windows install, how is going? Thanks.
Sincerely,
Yan
On Mon, Nov 27, 2017 at 2:14 PM, Mihai [email protected] wrote:
The errors on undefined identifiers would come either from missing CUDA dependencies or incompatible CUDA version
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/mihaibujanca/dynamicfusion/issues/33#issuecomment-347346589, or mute the thread https://github.com/notifications/unsubscribe-auth/AH0KC9qiGNJAC37CjM0dZUmX2Ozc1S4qks5s6zREgaJpZM4QlCbp .
Sorry. The last message is not correct. I saw all my device intrinsic functions defined by #if defined(CUDACC) are uncomment. That means __CUDACC__is not defined. But have no idea how to solve that. Google a lot and not find a solution.
Sincerely,
Yan
On Tue, Nov 28, 2017 at 2:32 PM, yan deng [email protected] wrote:
Hi Mihai,
I think the problem is here" Unsupported Features General CUDA ‣ CUDA library. The built-in functions __float2half_rn() and _half2float() have been removed. Use equivalent functionality in the updated fp16 header file from the CUDA toolkit." in https://developer.download. nvidia.com/compute/cuda/9.0/Prod/docs/sidebar/CUDA Toolkit_Release_Notes.pdf
Sincerely,
Yan
On Tue, Nov 28, 2017 at 2:22 PM, yan deng [email protected] wrote:
I use the latest CUDA version 9.0. Should I downgrade? First time using CUDA, lack of experience. Have you build this project on Windows successfully? If no one did, should I just give up and try other platform? I recalled you mentioned that you're preparing another tutorial for windows install, how is going? Thanks.
Sincerely,
Yan
On Mon, Nov 27, 2017 at 2:14 PM, Mihai [email protected] wrote:
The errors on undefined identifiers would come either from missing CUDA dependencies or incompatible CUDA version
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/mihaibujanca/dynamicfusion/issues/33#issuecomment-347346589, or mute the thread https://github.com/notifications/unsubscribe-auth/AH0KC9qiGNJAC37CjM0dZUmX2Ozc1S4qks5s6zREgaJpZM4QlCbp .
Ok, just got hold of a windows machine, will try building it soon. Are you using windows 10?
Yes. Thank you. I try to build everything with VS2015, 64bits.
Sincerely,
Yan
On Wed, Nov 29, 2017 at 4:56 AM, Mihai [email protected] wrote:
Ok, just got hold of a windows machine, will try building it soon. Are you using windows 10?
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/mihaibujanca/dynamicfusion/issues/33#issuecomment-347852320, or mute the thread https://github.com/notifications/unsubscribe-auth/AH0KC5bP1F9fB2DFRWz5xE7msGgqYIimks5s7VRmgaJpZM4QlCbp .
Hi Mihai,
I have found out reason why following error occurs. The .sln generated by cmake is not automatically linked to CUDA. Need to manually configure build dependencies/build customizations and choose cuda 9.0. And then configure all the properties related to CUDA C/C++ in property page. dynamicfusion\kfusion\src\cuda\device.hpp(54): error : identifier "__float2half_rn" is undefined dynamicfusion\kfusion\src\cuda\device.hpp(59): error : identifier "__half2float" is undefined dynamicfusion\kfusion\src\cuda\device.hpp(61): error : identifier "__half2float" is undefined dynamicfusion/kfusion/src/cuda/imgproc.cu(270): error : identifier "__float2half_rn" is undefined
Sincerely,
Yan
On Wed, Nov 29, 2017 at 10:13 AM, yan deng [email protected] wrote:
Yes. Thank you. I try to build everything with VS2015, 64bits.
Sincerely,
Yan
On Wed, Nov 29, 2017 at 4:56 AM, Mihai [email protected] wrote:
Ok, just got hold of a windows machine, will try building it soon. Are you using windows 10?
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/mihaibujanca/dynamicfusion/issues/33#issuecomment-347852320, or mute the thread https://github.com/notifications/unsubscribe-auth/AH0KC5bP1F9fB2DFRWz5xE7msGgqYIimks5s7VRmgaJpZM4QlCbp .
Any clue why would that happen?
How do you solve that in practice / is there a way I could add that to the repo so it happens automatically?
I suppose I could try building a .sln and include it in the repo, but I'm quite clumsy with developing on windows
http://docs.nvidia.com/cuda/cuda-installation-guide-microsoft-windows/index.html#build-customizations-for-existing-projects
I firstly did the customization following 3.4. Build Customizations for Existing Projects http://docs.nvidia.com/cuda/cuda-installation-guide-microsoft-windows/index.html#build-customizations-for-existing-projects so that you can have ncvv as compiler. Otherwise, all the intrinsic functions would not be defined. More specifically for VS2015, right click opened project, choose build dependencies/build customizations and select the CUDA.
Then in the property page, CUDA C/C++ would be available after add dependency, like following shows. I then configures all the entries in CUDA C/C++ and Linker as those provided by CUDA sample code, e.g., C:\ProgramData\NVIDIA Corporation\CUDA Samples\v9.0\5_Simulations\nbody. Details are listed as following. However, I'm not sure if they are all essential. I guess there should be some way to configure all those things in CMakeLists. I don't have experience with CUDA development and thus don't know how to do that. Will try to find answer after I build Opt.
The entries entered/modified in CUDA C/C++ are: common/additional include directories common/target machine platform device/code generation host/preprocessor definition host/runtime library command line/ additional options
The entries in Linker: general/additional library directiories input/additional dependencies.
[image: Inline image 1]
Sincerely,
Yan
On Wed, Nov 29, 2017 at 1:46 PM, Mihai [email protected] wrote:
Any clue why would that happen?
How do you solve that in practice / is there a way I could add that to the repo so it happens automatically?
I suppose I could try building a .sln and include it in the repo, but I'm quite clumsy with developing on windows
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/mihaibujanca/dynamicfusion/issues/33#issuecomment-348007039, or mute the thread https://github.com/notifications/unsubscribe-auth/AH0KC2CixI_nX7DMms9fprMfSJvOnb4Kks5s7dCwgaJpZM4QlCbp .
The following figure shows why functions would be shown as undefined if compiler nvcc is not linked. [image: Inline image 1]
On Wed, Nov 29, 2017 at 2:17 PM, yan deng [email protected] wrote:
http://docs.nvidia.com/cuda/cuda-installation-guide- microsoft-windows/index.html#build-customizations-for-existing-projects I firstly did the customization following 3.4. Build Customizations for Existing Projects http://docs.nvidia.com/cuda/cuda-installation-guide-microsoft-windows/index.html#build-customizations-for-existing-projects so that you can have ncvv as compiler. Otherwise, all the intrinsic functions would not be defined. More specifically for VS2015, right click opened project, choose build dependencies/build customizations and select the CUDA.
Then in the property page, CUDA C/C++ would be available after add dependency, like following shows. I then configures all the entries in CUDA C/C++ and Linker as those provided by CUDA sample code, e.g., C:\ProgramData\NVIDIA Corporation\CUDA Samples\v9.0\5_Simulations\nbody. Details are listed as following. However, I'm not sure if they are all essential. I guess there should be some way to configure all those things in CMakeLists. I don't have experience with CUDA development and thus don't know how to do that. Will try to find answer after I build Opt.
The entries entered/modified in CUDA C/C++ are: common/additional include directories common/target machine platform device/code generation host/preprocessor definition host/runtime library command line/ additional options
The entries in Linker: general/additional library directiories input/additional dependencies.
[image: Inline image 1]
Sincerely,
Yan
On Wed, Nov 29, 2017 at 1:46 PM, Mihai [email protected] wrote:
Any clue why would that happen?
How do you solve that in practice / is there a way I could add that to the repo so it happens automatically?
I suppose I could try building a .sln and include it in the repo, but I'm quite clumsy with developing on windows
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/mihaibujanca/dynamicfusion/issues/33#issuecomment-348007039, or mute the thread https://github.com/notifications/unsubscribe-auth/AH0KC2CixI_nX7DMms9fprMfSJvOnb4Kks5s7dCwgaJpZM4QlCbp .
I tried to use terra released version for windows. However, many problem occurs. The including "stdio,h" problem can be solved by add terralib.includepath = [[C:\Program Files (x86)\Windows Kits\10\Include\10.0.16299.0\ucrt;C:\Program Files (x86)\Microsoft Visual Studio 14.0\VC\include;C:\Program Files (x86)\Microsoft Visual Studio 14.0\VC\crt\src]] to the terra script. However, when saveobj is using, e.g., terralib.saveobj("hello",{main = hello}), terra is still not working. Unluckily, that function is required for Opt to be built. Changing terralib.saveobj("hello",{main = hello}) to terralib.saveobj("hello.o",{main = hello}) would generate hello.o. However, when using g++ to build executable file, error popped as following. error LNK2019: unresolved external symbol __acrt_iob_func referenced in function main error LNK2019: unresolved external symbol __stdio_common_vfprintf referenced in function _vfprintf_l I finally found out the reason is VS2015 https://community.microfocus.com/microfocus/mainframe_solutions/enterprise_developer_21/w/knowledge_base/27927/compatibility-issues-with-visual-studio-2015. However, I still don't know how to solve that. Changing to other VS might bring more problem since all the other libraries I built are based on VS2015, .
On Wed, Nov 29, 2017 at 2:25 PM, yan deng [email protected] wrote:
The following figure shows why functions would be shown as undefined if compiler nvcc is not linked. [image: Inline image 1]
On Wed, Nov 29, 2017 at 2:17 PM, yan deng [email protected] wrote:
http://docs.nvidia.com/cuda/cuda-installation-guide-microsof t-windows/index.html#build-customizations-for-existing-projects I firstly did the customization following 3.4. Build Customizations for Existing Projects http://docs.nvidia.com/cuda/cuda-installation-guide-microsoft-windows/index.html#build-customizations-for-existing-projects so that you can have ncvv as compiler. Otherwise, all the intrinsic functions would not be defined. More specifically for VS2015, right click opened project, choose build dependencies/build customizations and select the CUDA.
Then in the property page, CUDA C/C++ would be available after add dependency, like following shows. I then configures all the entries in CUDA C/C++ and Linker as those provided by CUDA sample code, e.g., C:\ProgramData\NVIDIA Corporation\CUDA Samples\v9.0\5_Simulations\nbody. Details are listed as following. However, I'm not sure if they are all essential. I guess there should be some way to configure all those things in CMakeLists. I don't have experience with CUDA development and thus don't know how to do that. Will try to find answer after I build Opt.
The entries entered/modified in CUDA C/C++ are: common/additional include directories common/target machine platform device/code generation host/preprocessor definition host/runtime library command line/ additional options
The entries in Linker: general/additional library directiories input/additional dependencies.
[image: Inline image 1]
Sincerely,
Yan
On Wed, Nov 29, 2017 at 1:46 PM, Mihai [email protected] wrote:
Any clue why would that happen?
How do you solve that in practice / is there a way I could add that to the repo so it happens automatically?
I suppose I could try building a .sln and include it in the repo, but I'm quite clumsy with developing on windows
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/mihaibujanca/dynamicfusion/issues/33#issuecomment-348007039, or mute the thread https://github.com/notifications/unsubscribe-auth/AH0KC2CixI_nX7DMms9fprMfSJvOnb4Kks5s7dCwgaJpZM4QlCbp .
I think that at this stage Ubuntu 16.04 should be the development environment for all to avoid losing time to support more platforms. Later when thirty party libraries will be better defined we should port it on Windows. I have written some instructions for compiling the project under Ubuntu, but now there are few changes like Opt instead of Ceres that I will update in instructions soon.
